Study population and setting
This study considered 24,808 patients (53% female) who were discharged from six hospitals within a large health care system in New York State from January to April 2020. The authors compared the occurrence of new-onset acute ischemic stroke between discharged patients with COVID-19 (with laboratory confirmed SARS-CoV-2 infection) and discharged patients without COVID-19, using a logistic regression model adjusted for age, sex, race, comorbidities, insurance status, and hospital. The outcome of stroke was restricted to those with symptoms and MRI findings consistent with acute ischemic stroke. Sensitivity analyses used propensity scores to address confounding variables.
Summary of Main Findings
During the study period, 2,513 patients (10.1%) tested positive for SARS-CoV-2 infection. Of these patients, 22 (0.9%) presented with acute ischemic stroke, and among the patients without SARS-CoV-2 infection, 544 (2.4%) presented with stroke. The mean age of those with stroke was 73 years. Those with COVID-19 had lower odds of stroke than other hospitalized patients; the adjusted odds ratio for stroke associated with COVID-19 was 0.25 (95% CI: 0.16 to 0.40). These results were similar in sensitivity analyses that used propensity scores to adjust for confounding. Those with stroke and COVID-19 had 10.5 times the odds of death compared to those with stroke but without COVID-19 (95% CI: 3.5 to 31.2).
The authors attempted to address possible bias by using propensity scores to weight observations, determined by a model including a broad range of comorbidities.
In observational studies like this one, there is a danger of inducing bias by selecting only hospitalized patients, which can create a non-causal association between variables. In this case, because patients could be hospitalized for COVID-19 or for other reasons, those without COVID-19 may have a higher risk of stroke as a consequence of selection, making it seem as if COVID-19 is protective for stroke (this phenomenon has been seen in previous studies of smoking and COVID-19). While propensity score weighting may be helpful in reducing this bias, it can only eliminate the bias if the mechanism for selection into the study is properly characterized, which is not assured in this case. Patients with severe respiratory distress on mechanical ventilation may not have been assessed for stroke, and coagulopathy was not necessarily recognized as an important sign of severe COVID-19 early in the pandemic, which could have led to an undercount of strokes or transient ischemic attacks in patients with COVID-19. This undermines the conclusion that stroke risk is lower among patients with COVID-19. The timing of stroke diagnoses was also not clear (e.g., whether stroke was only diagnosed at presentation to hospital or during admission). Finally, the biological plausibility of lower stroke risk in COVID-19 is questionable.
This study provides evidence that contradicts the prevailing impression of increased stroke risk among patients with COVID-19.
This review was posted on: 5 December 2020